Viral delivery of shRNA to amygdala neurons leads to neurotoxicity and deficits in Pavlovian fear conditioning

The use of viral vector technology to deliver short hairpin RNAs (shRNAs) to cells of the nervous system of many model organisms has been widely utilized by neuroscientists to study the influence of genes on behavior. However, there have been numerous reports that delivering shRNAs to the nervous system can lead to neurotoxicity. Here we report the results of a series of experiments where adeno-associated viruses (MV), that were engineered to express shRNAs designed to target known plasticity associated genes (i.e. Arc, Egr1 and GluN2A) or control shRNAs that were designed not to target any rat gene product for depletion, were delivered to the rat basal and lateral nuclei of the amygdala (BLA), and auditory Pavlovian fear conditioning was examined. In our first set of experiments we found that animals that received MV (3.16E13-1E13 GC/mL; 1 mu l/side), designed to knockdown Arc (shArc), or control shRNAs targeting either luciferase (shLuc), or nothing (shCntrI), exhibited impaired fear conditioning compared to animals that received viruses that did not express shRNAs. Notably, animals that received shArc did not exhibit differences in fear conditioning compared to animals that received control shRNAs despite gene knockdown of Arc. Viruses designed to harbor shRNAs did not induce obvious morphological changes to the cells/tissue of the BLA at any dose of virus tested, but at the highest dose of shRNA virus examined (3.16E13-1E13 GC/mL; 1 mu l/side), a significant increase in microglia activation occurred as measured by an increase in IBA1 immunoreactivity. In our final set of experiments we infused viruses into the BLA at a titer of (1.60E+12 GC/mL; 1 mu l/side), designed to express shArc, shLuc, shCntrl or shRNAs designed to target Egr1 (shEgr1), or GluN2A (shGluN2A), or no shRNA, and found that all groups exhibited impaired fear conditioning compared to the group which received a virus that did not express an shRNA. The shEgr1 and shGluN2A groups exhibited gene knockdown of Egr1 and GluN2A compared to the other groups examined respectively, but Arc was not knocked down in the shArc group under these conditions. Differences in fear conditioning among the shLuc, shCntrl, shArc and shEgr1 groups were not detected under these circumstances; however, the shGluN2A group exhibited significantly impaired fear conditioning compared to most of the groups, indicating that gene specific deficits in fear conditioning could be observed utilizing viral mediated delivery of shRNA. Collectively, these data indicate that viral mediated shRNA expression was toxic to neurons in vivo, under all viral titers examined and this toxicity in some cases may be masking gene specific changes in learning. Therefore, the use of this technology in behavioral neuroscience warrants a heightened level of careful consideration and potential methods to alleviate shRNA induced toxicity are discussed. (C) 2015 Elsevier Inc. All rights reserved.